What’s behind AlphaFold’s latest AI breakthrough?
When my phone rang yesterday afternoon I wasn’t going to answer it. It’s like when you don’t want to be woken up because you know you’ll never get back to the dream you were having. I was working on a coding problem, retaining dozens of different threads in my mind of potential solutions: “If this bit works then next I’ll try this… but if the first bit doesn’t work, I’ll try this instead… or if the first and second bits work that could help the area I worked on yesterday to be better because…” You get the idea.
Against my better judgement I took the call, feeling the creativity, ideas and inspiration draining from my brain as I did so.
My friend Angela is not known for being excitable so the waves of euphoria coming down the line from her were very surprising.
“Have you seen the news? Your lot have done it!” she gasped.
“Errm, my lot? Who? Done what?”
“Haven’t you seen the news?”
Well, I’d been reading, writing and coding all day so I tend to ignore the news as its mostly reconstituted tedium that I’ll catch up on later in the day. Just because we have 24 hour news doesn’t mean there’s anything newsworthy going on normally so what had I missed? I guessed it must be something to do with Artificial Intelligence (AI) because, since I started studying for a Masters degree in AI and Data Science my friends treat me like I’ve joined some obscure tribe and am therefore intimately connected to every AI announcement (good or bad) that they read.
Angela explained that “my lot” was DeepMind who had announced that they had found a way to use AI to predict how proteins fold, which has a major impact on how these vital building blocks work. A pattern that can take a researcher years to find could now be uncovered by a computer in a day.
Angela is a biologist so this discovery meant a lot to her and, as I hastily read the DeepMind press release, she waxed lyrical about what solving the protein folding problem could mean for her discipline and what else DeepMind might use their skills to do next to advance biology.
My career to date has been in communications/PR , marketing and government relations so I know not to take any press release at face value. I didn’t want to pour cold water on Angela’s enthusiasm, but the sceptic in me always looks for the angle behind the story.
“Stop a minute,” I said. “I can see why this means so much to biologists, medical researchers and many other fields, but I wouldn’t be surprised if DeepMind now shift their focus onto something completely different — like predicting the trajectory of asteroids that might crash into earth with the resulting cloud of ash and steam plunging us into a centuries long, crop-killing, species annihilating dark winter.”
Now it was Angela’s turn to be flummoxed. “What? Why would they do that after they’ve cracked this 50 year mystery when there’s so much more to do to develop it into practical tools?”
“Because DeepMind isn’t about solving particular problems; the company wants to create superintelligence so they find problems that human brains are struggling to crack and see if they can use AI to solve them. The point for them is not what problem they solve, but what they learn about thinking and computing from finding the solution. They said it themselves in their blog yesterday:
‘This breakthrough demonstrates the impact AI can have on scientific discovery and its potential to dramatically accelerate progress in some of the most fundamental fields that explain and shape our world.’
“There are scores of teams at DeepMind and other companies round the world learning how to learn, how to crack the world’s biggest conundrums using AI so who knows what they will discover next? But it’s a fair bet that it will have nothing to do with biology.”
Angela and I ended the call agreeing that it wasn’t ethically right for all the world’s attention to be put on solving just one problem so it was fair for DeepMind to shift its resources now onto another big challenge and for others to take up the mantle of improving protein folding prediction.
I then went back to my screen….and bashed my forehead against my desk for a minute or two when I realised that the brilliant idea I’d had 20 minutes earlier had now vanished. Superintelligence wouldn’t forget so easily.
